2026 Cryptocurrency Investment Outlook: Rise of Application Chains, AI Agents Take Over DeFi

Author: Archetype Compiled by: Shen Chao TechFlow Link:

Disclaimer: This article is a reprint. Readers can obtain more information through the original link. If the author has any objections to the reprint, please contact us, and we will make modifications according to the author’s requirements. Reprints are for information sharing only, do not constitute any investment advice, and do not represent Wu Shuo’s views and positions.

The Era of Building Dedicated Application Chains Has Finally Arrived

Author: Aadharsh Pannirselvam

In simple terms: Blockchains designed, built, and optimized for applications will bring disruptive change. The best application chains next year will focus on core components and fundamental principles, deliberately and carefully assembled.

Recently, developers, users, institutions, and capital flooding onto chains are different from before: they pay more attention to specific cultures (i.e., the definition of user experience) rather than pursuing abstract ideals such as decentralization or censorship resistance. In practice, this demand sometimes aligns with existing infrastructure, and sometimes it does not.

For user-oriented abstract applications like Blackbird or Farcaster, certain centralized design decisions—such as node co-location, a single sequencer, and custom databases—might have been considered heretical three years ago, but now they seem very reasonable. The same applies to stablecoin chains and trading platforms like Hyperliquid* and GTE, whose success often depends on millisecond-level latency, price volatility, and optimal pricing.

However, this logic does not apply to all new applications.

For example, the increasing focus on privacy by institutions and retail users is opposite to this acceptance of centralization. The demands of crypto applications and ideal user experiences can differ greatly, and their infrastructure should also vary accordingly.

Fortunately, building chains from scratch to meet these specific user experience definitions is no longer as complex as it was two years ago. In fact, the process is very similar to assembling a custom PC.

Of course, you can choose each hard drive, fan, and cable yourself. But if you don’t need that level of detailed customization (which is likely the case for most people), you can use services like Digital Storm or Framework, which offer a range of pre-built custom computers to meet different needs. If your requirements are between these two extremes, you can also add your own components on top of the already selected and compatible parts. This approach not only provides higher modularity and flexibility but also saves unnecessary components while ensuring the final product runs efficiently.

When assembling and tuning core components such as consensus mechanisms, execution layers, data storage, and liquidity, various applications are creating forms with cultural uniqueness, continuously reflecting different needs (i.e., different definitions of user experience), serving their specific target audiences, and ultimately preserving value. These differences may be as significant as the distinctions among rugged ToughBooks, business-oriented ThinkPads, powerful desktops, and elegantly designed MacBooks, but they also tend to converge and coexist to some extent—after all, not every one of these computers runs its own independent operating system. Moreover, each essential component becomes a “knob” that applications can flexibly adjust, allowing developers to optimize at will without worrying about destructive changes to the underlying protocol.

Considering Circle’s acquisition of Malachite from Informal Systems, it’s clear that having control over customized blockchain space has now become a broader priority. In the coming year, I look forward to seeing applications and teams define and control their chain resources around core components and sensible defaults provided by companies like Commonware and Delta. This model is similar to HashiCorp or Stripe Atlas but applied to blockchain and the blockchain space.

Ultimately, this will enable applications to directly control their cash flows and leverage the uniqueness of their construction to deliver the best user experience under their own conditions, forming a lasting competitive barrier.

Prediction Markets Will Continue to Innovate (But Only Some Will Succeed)

Author: Tommy Hang

In this cycle, prediction markets have become one of the most watched applications. With all crypto sector’s daily trading volume hitting a record high of $2 billion, this category has clearly taken a significant step toward mainstream consumer products.

This momentum has benefited a series of related projects that attempt to supplement or challenge current market leaders like Polymarket and Kalshi. However, distinguishing genuine innovation from noise in this wave ultimately determines which projects we should continue to follow into 2026.

From a market structure perspective, I especially look forward to solutions that can narrow spreads and deepen open interest. Although market creation remains permissioned and selective, liquidity in prediction markets is still relatively low for market makers and traders. There are significant opportunities to improve optimal routing systems, different liquidity models, and collateral efficiency through lending and other products.

Trading volume by category is also a key factor in determining which platforms will win. For example, over 90% of Kalshi’s November trading volume came from sports markets, indicating that some platforms have a natural advantage in competing for liquidity. In contrast, Polymarket’s trading volume in crypto-related and political markets is 5 to 10 times higher than Kalshi’s.

However, on-chain prediction markets still have a long way to go before achieving true mass adoption. A good benchmark is the Super Bowl in 2025, which alone generated $23 billion in on-chain betting volume in one day—more than ten times the total daily trading volume of all existing on-chain markets.

To close this gap, sharp and inspiring teams need to solve core issues in prediction markets. Over the next year, I will closely watch these potential industry players.

Smart Agent Curators Will Drive DeFi Expansion

Author: Eskender Abebe

Currently, the curation layer in DeFi exists in two extreme forms: fully algorithmic (hardcoded interest rate curves, fixed rebalancing rules) and fully manual (risk committees, active managers). Smart agent curators (Agentic Curators) represent a third mode: AI agents—including large language models (LLMs) + tools + loops—manage curation and risk policies in treasuries, lending markets, and structured products. This is not just executing fixed rules but involves reasoning and decision-making about risk, returns, and strategies.

An example is the role of curators in Morpho markets. Here, someone needs to define collateral policies, loan-to-value (LTV) limits, and risk parameters to generate yield products. Currently, human factors in this process are a bottleneck. Smart agents can scale this process. Soon, smart agent curators will directly compete with algorithmic models and human managers.

So, when will the “Move 37” moment in DeFi arrive?

When I talk to crypto fund managers about AI, they usually give two very different answers: either they believe LLMs will soon automate every trading desk, or they think these tools are just “illusory toys” that cannot handle real markets. But both overlook a key architectural shift: smart agents can introduce emotionless execution, systematic policy adherence, and flexible reasoning into areas where human performance is noisy, pure algorithms are too fragile, or both.

They are likely to supervise or combine lower-level algorithms rather than fully replace them. In this scenario, LLMs are more like “architects” designing safety frameworks, while deterministic code continues to handle the latency-sensitive core paths.

When deep reasoning costs drop to a few cents, the most profitable treasuries will no longer be those with the smartest humans but those with the most computational power.

Short-Form Video Becomes the New “Storefront”

Author: Katie Chiou

Short-form videos are rapidly becoming the default interface for discovering (and ultimately purchasing) content you love. TikTok Shop achieved over $20 billion GMV in the first half of 2025, nearly doubling year-over-year, quietly transforming global entertainment consumption into a “storefront” experience.

In response, Instagram has turned Reels from a defensive feature into a revenue engine. This short-video format not only brings more exposure but is also capturing an increasing share of Meta’s projected ad revenue in 2025. Meanwhile, Whatnot has demonstrated that live-streamed, personality-driven sales conversions surpass traditional e-commerce.

The core logic is simple: people make decisions faster when watching content in real time. Every swipe becomes a decision point. Major platforms understand this well, and the boundary between recommendation feeds and checkout processes is disappearing. Information streams are becoming new sales channels, and every creator is a distribution channel.

Artificial intelligence further accelerates this trend. AI reduces video production costs, increases content output, and enables creators and brands to test new ideas in real time. More content means more conversion opportunities, and platforms optimize each second of video to enhance purchase intent.

Cryptocurrency plays a key role in this trend. Faster content requires faster, more efficient payment rails. As shopping becomes seamless and embedded directly into content, we need a system capable of handling micro-payments, programmatic revenue sharing, and tracking contributions across complex influence chains. Cryptocurrency is born for this liquidity. It’s hard to imagine a large-scale, media-native commercial era without crypto support.

Blockchain Will Drive New Rules for AI Expansion

Author: Danny Sursock

In recent years, AI has focused on a multi-billion-dollar arms race among mega-corporations and startups, while decentralized innovators have been quietly exploring in the shadows.

However, as mainstream attention shifts, some crypto-native teams have made significant progress in decentralized training and inference. This quiet revolution is gradually moving from theory to testing and production.

Today, teams like Ritual*, Pluralis, Exo*, Odyn, Ambient, Bagel, and others are preparing for the spotlight. These new competitors are poised to unleash explosive disruptive impacts on the foundational development trajectory of AI.

By training models in globally distributed environments, leveraging new asynchronous communication and parallelization methods validated at production scale, the limits of AI expansion will be thoroughly broken.

Meanwhile, the combination of new consensus mechanisms and privacy primitives makes verifiable and confidential inference a practical option for on-chain developer tools.

Furthermore, revolutionary blockchain architectures will truly realize the integration of smart contracts with highly expressive computational structures, which can run autonomous AI agents using cryptocurrency as a medium of exchange, simplifying their operation.

The foundational work is already done.

The current challenge is to scale these infrastructures to production levels and demonstrate that blockchain can drive fundamental AI innovations beyond philosophical, ideological, or superficial funding experiments.

实体资产将迎来真正的链上普及

作者：Dmitriy Berenzon

多年來，我們一直在討論資產通證化。然而，隨著穩定幣的主流採用、順暢且強大的法幣與加密貨幣的出入金通道的出現，以及全球範圍內日益明確的監管支持，實體資產（Real World Assets，RWAs）終於開始實現大規模應用。根據 RWA.xyz* 的數據，截至目前，各類別的通證化資產總額已超過 180 億美元，而一年前這一數字僅為 37 億美元。我預計，到 2026 年，這一趨勢將進一步加速。

需要注意的是，通證化和金庫（Vaults）是 RWAs 的兩種不同設計模式：通證化是將鏈下資產的表示形式搬到鏈上，而金庫則是搭建鏈上資本與鏈下收益之間的橋梁。

我很期待看到通證化和金庫能夠為各種實體和金融資產提供鏈上接入，從黃金和稀有金屬等大宗商品，到用於營運資金和支付融資的私人信貸，再到私募和公開股票，以及更多的全球貨幣。當然，我們也可以大膽創新！我希望看到雞蛋、GPU、能源衍生品、工資預支、巴西國債、日元等資產都搬到鏈上！

但需要明確的是，這不僅僅是為了把更多東西放到鏈上，而是為了通過公共區塊鏈升級全球資本配置方式。區塊鏈能讓那些模糊、緩慢且孤立的市場變得透明、可編程且流動性更強。而一旦這些資產被搬到鏈上，我們就能享受到與現有 DeFi 原語的可組合性所帶來的巨大優勢。

最後，許多這些資產在鏈上化的過程中不可避免地會面臨可轉讓性、透明度、流動性、風險管理和分發方面的挑戰，因此能夠緩解這些問題的基礎設施建設同樣重要且令人興奮！

代理驅動的產品復興即將到來

作者：Ash Egan

未來的網絡形態將更少由我們滑動的社交平台決定，而更多由我們對話的智能代理（agents）塑造。

如今，機器人和智能代理已經佔據了網絡活動中快速增長的一部分。根據粗略估算，這一比例目前約為 50%，包括鏈上和鏈下的活動。在加密領域，機器人正越來越多地參與交易、策展、協助、掃描合約，並在如代幣交易、資金管理、智能合約審計和遊戲開發等方面為我們代理執行任務。

這標誌著一個可編程、代理驅動的網絡時代的到來。雖然我們已經處於這個階段有一段時間了，但 2026 年將成為一個轉折點，加密產品設計將更多地服務於智能代理，而非直接面向人類（以一種積極、解放且非反烏托邦的方式）。

這種未來的圖景正在逐漸成型。我個人希望能夠減少在不同網站間點擊的時間，而更多地通過類似聊天界面的方式，管理鏈上的智能代理。想像一下，像 Telegram 一樣的界面，但對話對象是針對特定應用或任務的智能代理。這些代理能夠制定並執行複雜的策略，搜索網絡中與我最相關的信息和數據，並反饋交易結果、風險、機會以及經過篩選的信息。我只需要給出任務，它們就會找到機會，過濾掉噪音，並在最佳時機執行操作。

鏈上的基礎設施已經為此做好了準備。將默認開放的數據圖譜、可編程的微支付、鏈上社交圖譜以及跨鏈流動性軌道結合起來，我們已經擁有支持一個動態智能代理生態系統所需的一切。加密領域的即插即用特性意味著代理在這裡面臨更少的繁瑣流程和無效路徑。與 Web2 的基礎設施相比，區塊鏈為這種智能代理化提供了絕佳的條件。

這可能是最重要的一點：這不僅僅是自動化，而是從 Web2 的孤島中解放出來。從摩擦中解放出來。從等待中解放出來。我們在搜索領域看到這種轉變的發生：目前約 20% 的 Google 搜索會生成 AI 概覽，而數據顯示，當人們看到 AI 概覽時，他們點擊傳統搜索結果鏈接的概率顯著降低。手動翻閱頁面變得不再必要。可編程的智能代理網絡將進一步擴展這種趨勢，覆蓋到我們使用的應用程序，而我認為這是一個積極的變化。

這一時代將讓我們減少無意義的刷屏，減少恐慌性交易。時區差異將逐漸消失（不再需要“等亞洲市場醒來”）。與鏈上世界的互動將變得更加簡單且富有表現力，無論是對於開發者還是普通用戶。

隨著更多的資產、系統和用戶逐步進入鏈上，這一循環將不斷放大：

更多鏈上機會 → 部署更多智能代理 → 解鎖更多價值。循環往復。

但我們現在所構建的內容，以及構建的方式，將決定這個智能代理網絡是成為一層薄弱的噪音與自動化，還是點燃一場賦能且充滿活力的產品復興。

*註：文中提及的部分公司為 Archetype 投資組合公司。